Methodology of Control

The ultimate goal of science is and ought to be, to offer the method of improving control over the external world, to pave the way from the present state of affairs to the most preferred one. The field of international relations must not be an exception, as time, effort, and money invested in the research are useful as long as its intellectual product helps achieve the goals of actors such as government agencies, transnational corporations, NGOs, and others. If we agree on that (the following discussion will not make sense unless we do), then a fundamental question comes to emerge, namely: “what scientific tools may be used to improve control over processes in international relations?” I shall provide the answer in three steps by consequently defining the concepts of (1) international relations and (2) intervention, and figuring out (3) which methodology is better for our purposes.

WHAT IS “INTERNATIONAL RELATIONS”?

Obviously, there can be no relations between states, countries, or nations, since the latter are aggregates, and literal interpretation of the term “international relations” attaches to the aggregates the properties of their elements, which is the fallacy of composition. Consequently, conflicts, negotiations, summits etc. should rather be viewed as relations between individuals. Secondly, it seems correct to replace ‘relations’ with ‘interactions’ because the former implies a significant introspective component, not subject to observation, whereas ‘interactions’ is free from such flaws and merely means interpersonal behavior. Thirdly, the interactions we are interested in are those between residents of different states, since this is what follows from the metaphor of ‘international relations’. Finally, these interactions are expected to cause redistribution of economic goods (international economic relations) or reshaping spheres of influence (international [political] relations).

WHAT IS CONTROL?

Vaguely put, control is the ability to ensure the occurrence of some events and to prevent that of others. As the concept of control implies the subject and the object (say, the US President and the Assad regime), determining the ability of control largely boils down to revealing causal relationships between what the subject can do and the effect of these efforts. Such causation may be easily demonstrated through the notions of necessary and sufficient conditions:

1.2. military intervention (P) is a sufficient condition for toppling the Assad regime down (Q) if P → Q, that is, the realization of P necessities Q

1.2. intervention is a necessary condition for regime change if ¬P → ¬Q, that is, if P does not realize, Q does not either;

Despite its undoubted appeal, the instrument can hardly be used for real-life situations where same policies may lead to different outcomes which are determined by a number of factors, so sufficient and necessary condition had better be replaced with the notions of enhancing and inhibiting factors:

1.3. military intervention (F) is an enhancing factor of toppling the Assad regime down (E), if P(E|F) > P(E), that is, the probability of E increases with the realization of F;

1.4. intervention is an inhibiting factor of toppling the regime if P(E|F) < P(E), that is, the realization of F decreases the probability of E;

However, this model of causation is deemed unsatisfactory, too, since it does not take account of the intensity of causes and effects. For instance, military intervention may realize as the deployment of a single battalion, or of several formations, and the difference must not be neglected. Therefore, let us switch our attention to mathematics and regard causal relationships as a function:

2.1. y = f(x, y, …, n), where y – dependent variable, arguments x, y, …, n – explanatory variables;

To give an example, the causal relationship between two variables may be presented by the linear function:

2.2. y = ax + b, where x – explanatory variable, a and b – some parameters; If the dependent variable is categorical (regime falls/regime survives), the most appropriate solution is logistic regression:

2.3.

Despite a larger degree of sophistication, mathematical and statistical models are rather helpless at distinguishing between causes and effects because the correlation between variables may very well be the influence of the common cause, the result of mutual determination, a mere coincidence, or the product analytical relationships. That is why the choice of methodology primarily hinges upon its ability to reveal causal relationship through both logical analysis (1.1 – 1.4) and statistics (2.1-2.3).

WHICH METHODOLOGY IS BETTER?

Let us consider main approaches used in social sciences:

• behaviorism
• rational choice (game theory, public choice etc.)
• hermeneutics-based theories (cultural studies, feminism, and so on)
• holism-based theories (systems thinking, structuralism, Marxism, and others)

The worst choice would be holism. Firstly, it is barely compatible with statistics as selecting political systems as units of analysis does not allow to have data arrays of more than fifty-odd cases which is why non-existing correlation is likely to emerge. Secondly, the very idea of holism that the whole is larger than the sum of its parts is an unjustified assumption rather than an empirically verifiable statement. Although in case of human psyche, as Gestalt psychology asserts, perception of the whole does not indeed boil down to the sum of perceptions of its parts, in social sciences holism clearly lays an ontological claim whose absurdity has been demonstrated by logical positivists. Any meaningful statement must relate to empirical facts, therefore the meaning of an abstract concept by definition consists of the facts that its parts relate to. For example, “society” can thus serve only as a convenient tool to denote individuals who comprise it, whereas any claims for a broader sense are no other than commonplace mysticism. And thirdly, interpretation of conclusions about large units of analysis is inevitably flawed with the fallacy of division. If, say, Egypt’s political system is predicted to stagnate or degrade within the following five-year period, does it mean that every part of it is going to? Obviously not, but such interpretation is implied by holistic theories so they overall fail at putting forward and verifying meaningful hypotheses.

Hermeneutics-based theories are not necessarily incompatible with statistical analysis as a huge part of cultural studies are carried out at the individual level with the help of large survey data, but are their conclusions useful at all? It is true that individuals’ behavior seems to be determined by their mental dispositions, that is, I am writing this not exclusively because of the beneficial consequences expected to improve my utility level, but also because I generally like writing on science and philosophy and see some internal value in it. However, this explanation is trivially right: free decisions by individuals merely reflect their values which affords an opportunity to rationalize any behavior such as “the Iraq War has been caused by the hawkishness of the Bush administration” or “Russia’s political ‘thaw’ during 2008-2012 is explained by D. Medvedev’s liberal beliefs”. But, unlike past behavior which can always be said to be determined by the agent’s values, predicting future developments are not an easy task for hermeneutics because we would then have to calculate all changes over time in numerous mental characteristics which is fairly impossible. Finally, the empirical test of hypotheses about value-determined behavior demands an objective source of data on the individual’s values which science clearly does not possess. Deducting values from behavior generates trivial conclusions demonstrated above whereas the use of introspection (for instance, (wide) reflective equilibrium by John Rawls and Norman Daniels or psychoanalysis) deals with the individuals judgments (often erroneous) on their values, not values themselves. Therefore, hermeneutics-based theories are ill suitable for our purposes, too.

The rest are behaviorism and rational choice, both of which may be reconstructed by forming two sets of assumptions with elements of one obtained by negating those of the other. Since close examination of the two methodologies takes a whole new entry rather than a couple of paragraphs, I shall continue to solve the problem outlined next time.